[DocDB] Remove DocKey decoding part from ScanChoices

[karan-yb]

Jira Link: DB-4457

Description

As part of the fix for #12899 and #14862, we figured that ScanChoices SkipTargetsUpTo can be optimized further by avoiding the decode for key columns which are not used in the filter. Based on the offline discussion, here is the proposal discussed in meeting:

1. ScanChoices - introduce a new method which returns the filtered key column indexes.
   std::vector<size_t> GetFilteredKeyColumnIndexes()
2. Iterator - decode the key before calling SkipTargetsUpTo and handle system keys as part of it. Update ScanChoices SkipTargetsUpTo to:
   Status SkipTargetsUpTo(Slice key_prefix, std::vector<KeyEntryValue or Slice>& key_columns_or_slice, Slice suffix)

• where key_prefix is the part of the key which includes hash key and other metadata. It also includes the key columns which are not required by filter.
• key_columns_or_slice - vector either includes a decoded KeyEntryValue for a filtered column or Slice for key column which is not required by ScanChoices
• suffix includes all key columns which are not part of filter